#16232运行onnx模型推理时，报错"invalid input 'img' in NormalizeImage"

Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 80+ languages.

Issue Details

2 months ago

Bobholamovic

status/close

View on GitHub

lightlpr

opened 2 months ago

Author

🔎 Search before asking

I have searched the PaddleOCR Docs and found no similar bug report.
I have searched the PaddleOCR Issues and found no similar bug report.
I have searched the PaddleOCR Discussions and found no similar bug report.

🐛 Bug (问题描述)

按照在线文档 http://www.paddleocr.ai/latest/version2.x/legacy/paddle2onnx.html?h=onnx#_3 文档里给的步骤，在release/3.1版本上运行PP-OCRv4_mobile_det_infer， PP-OCRv4_mobile_rec_infer和ch_ppocr_mobile_v2.0_cls_infer，报下面的错误，不知道3.1版本是否支持onnx推理？

$ python3 tools/infer/predict_system.py --use_gpu=False --use_onnx=True --det_model_dir=./inference/det_onnx/model.onnx --rec_model_dir=./inference/rec_onnx/model.onnx --cls_model_dir=./inference/cls_onnx/model.onnx --image_dir=./docs/images/demo.gif
[2025/08/14 08:50:15] ppocr INFO: In PP-OCRv3, rec_image_shape parameter defaults to '3, 48, 320', if you are using recognition model with PP-OCRv2 or an older version, please set --rec_image_shape='3,32,320
<ppocr.data.imaug.operators.DetResizeForTest object at 0x7f4d86e74ee0>
<ppocr.data.imaug.operators.NormalizeImage object at 0x7f4d86e74f70>
image shape----- (512, 960, 3) <class 'numpy.ndarray'>  Normalize shape  0.003921569 [[[0.485 0.456 0.406]]] (1, 1, 3)
Traceback (most recent call last):
  File "/train-data/research/PaddleOCR/tools/infer/predict_system.py", line 326, in <module>
    main(args)
  File "/train-data/research/PaddleOCR/tools/infer/predict_system.py", line 227, in main
    dt_boxes, rec_res, time_dict = text_sys(img)
  File "/train-data/research/PaddleOCR/tools/infer/predict_system.py", line 109, in __call__
    dt_boxes, elapse = self.text_detector(img)
  File "/train-data/research/PaddleOCR/tools/infer/predict_det.py", line 407, in __call__
    dt_boxes, elapse = self.predict(img)
  File "/train-data/research/PaddleOCR/tools/infer/predict_det.py", line 250, in predict
    data = transform(data, self.preprocess_op)
  File "/train-data/research/PaddleOCR/ppocr/data/imaug/__init__.py", line 74, in transform
    data = op(data)
  File "/train-data/research/PaddleOCR/ppocr/data/imaug/operators.py", line 84, in __call__
    assert isinstance(img, np.ndarray), "invalid input 'img' in NormalizeImage"
AssertionError: invalid input 'img' in NormalizeImage

🏃‍♂️ Environment (运行环境)

OS ubuntu18, conda
python 3.9.13
PaddleOCR 3.1

🌰 Minimal Reproducible Example (最小可复现问题的Demo)

python3 tools/infer/predict_system.py --use_gpu=False --use_onnx=True --det_model_dir=./inference/det_onnx/model.onnx --rec_model_dir=./inference/rec_onnx/model.onnx --cls_model_dir=./inference/cls_onnx/model.onnx --image_dir=./docs/images/demo.gif

PaddleOCR

PaddlePaddle